2c3672f2cb
- AdaptiveGate rewritten (Netflix Gradient2): window-vs-window per-stream speed gradient, count-driven windows with admission stamps, no thresholds or mode state machine; failures x0.7 + 30s drain pause - session-level admission for multi-turn agents (_SessionGate): in-progress sessions hold slots until done, newcomers queue at the door; capacity follows the model gate's discovered limit (CONCUR-style continuity) - image service: memory-first register (zero docker calls for known images), TTL-cached docker images listing, optimistic ready when the daemon is unreachable (docker save contention no longer kills runs); es tar loading removed in favor of ModelScope shipping (ms_images.py per-image tar upload/pull with round-trip verification) - runner: circuit breaker (12 consecutive failures abort the bench), first-failure error printed immediately - swe_agentic: image wait / docker run / rm off the event loop; exec timeout becomes an observation the agent can react to; container gets curlrc + git low-speed aborts (stalled github downloads fail fast) - eval run excludes its own endpoints from http_proxy (a sick personal proxy read as 'endpoint dead' and killed whole runs) - progress bar shows failed count; swe agentic exec_workers 2 -> 4 Co-Authored-By: Claude <noreply@anthropic.com>
75 lines
1.6 KiB
YAML
75 lines
1.6 KiB
YAML
default:
|
||
temperature: 0.0
|
||
top_p: 1.0
|
||
stream: true
|
||
max_tokens: 32768
|
||
max_turns: 200
|
||
aime24:
|
||
temperature: 1.0
|
||
repeats: 12
|
||
max_tokens: 8192
|
||
aime25:
|
||
temperature: 1.0
|
||
repeats: 12
|
||
max_tokens: 8192
|
||
aime26:
|
||
temperature: 1.0
|
||
repeats: 12
|
||
max_tokens: 8192
|
||
hmmt26:
|
||
temperature: 1.0
|
||
repeats: 12
|
||
max_tokens: 8192
|
||
imo_answerbench:
|
||
temperature: 1.0
|
||
gpqa_diamond:
|
||
temperature: 1.0
|
||
repeats: 3
|
||
max_tokens: 8192
|
||
mmlu:
|
||
max_tokens: 8192
|
||
mmlu_pro:
|
||
max_tokens: 8192
|
||
cmmlu:
|
||
max_tokens: 8192
|
||
arc:
|
||
max_tokens: 8192
|
||
hellaswag:
|
||
max_tokens: 8192
|
||
winogrande:
|
||
max_tokens: 8192
|
||
simple_qa:
|
||
max_tokens: 8192
|
||
trivia_qa:
|
||
max_tokens: 8192
|
||
humaneval:
|
||
temperature: 1.0
|
||
repeats: 3
|
||
live_code_bench:
|
||
temperature: 1.0
|
||
longbench_v2:
|
||
max_tokens: 8192
|
||
max_input_tokens: 128000
|
||
openai_mrcr:
|
||
max_tokens: 8192
|
||
max_input_tokens: 128000
|
||
bfcl_v3:
|
||
max_tokens: 4096
|
||
env: bfcl_mock # agent 模式：工具调用轨迹 + 官方 AST 判分
|
||
general_fc:
|
||
max_tokens: 4096
|
||
tau2_bench:
|
||
max_tokens: 16384
|
||
env: tau2_official # agent 模式：官方引擎
|
||
max_turns: 200 # 一次完整电话流程（问候→验证→找单→政策→执行→确认）
|
||
|
||
swe_bench_verified:
|
||
max_tokens: 32768 # patch 生成预算（diff 可能较长）
|
||
temperature: 0.0 # es 口径：确定性生成 patch
|
||
|
||
swe_bench_verified_agentic:
|
||
max_tokens: 4096 # 单轮 bash 命令生成预算（mini-swe-agent 口径）
|
||
temperature: 0.0
|
||
env: swe_agentic # 多轮 agent：bash 探索 /testbed + sentinel 提交
|
||
max_turns: 100 # mini-swe-agent 默认步数
|